Anthropic would need about $602 billion in annual revenue to earn a hypothetical 20% return on roughly $1 trillion of assumed invested capital, Financial Times columnist Robert Armstrong calculates. He’s modeling what financial success might require over roughly five to 10 years, not reporting an Anthropic revenue target. Source
How the calculation works
In the FT’s Unhedged opinion newsletter, Armstrong cites Anthropic’s S-1 filing for approximately $901 billion in off-balance-sheet financial commitments, substantially related to compute and infrastructure, plus additional commitments estimated at $150 billion to $175 billion. Source
He rounds that to $1 trillion and assumes, for the exercise, that it represents all the capital that will sit inside the business. That’s a modeling choice, not a statement that Anthropic has already spent $1 trillion—or that future commitments and invested capital are interchangeable accounting measures. Armstrong says the commitments extend over about 15 years; he cites Zion Research Group’s estimate that $700 billion comes due within five. Source
His model then assumes:
- 20% return on invested capital: After-tax operating profit divided by invested capital. On the assumed $1 trillion capital base, that requires $200 billion a year in after-tax operating profit.
- 40% operating margin: Operating profit would equal 40% of revenue.
- 17% effective tax rate: Accounting for tax brings the required annual revenue to approximately $602 billion. Source
Neither the operating margin nor the tax rate is presented as Anthropic’s actual performance. Change the assumed capital base, return benchmark, margin or tax rate, and the revenue requirement changes too. Source
An Apple-sized comparison, and then some
For scale, Armstrong lists annual revenue of $478 billion at Apple and $471 billion at Alphabet. His hypothetical Anthropic figure exceeds either company’s sales and approaches two-thirds of the $908 billion he lists for all 76 software and IT services companies in the S&P 1500. Even this thought experiment needs a very large checkout counter. Source
The 20% benchmark isn’t a promised return on someone’s shares, and the calculation doesn’t establish what happens if Anthropic falls short. Source
Armstrong closes with a question he credits to Richard Kramer of Arete Research: “where is all that revenue going to come from?” Source
Comments